Many on Twitter - though not
itself - are calling YouGov's Scotland survey/independence referendum prediction (released tonight) an "exit" poll (examples below). It is not. No judgment here on the survey itself, just a quick and likely-futile-but-I'll-bang-my-head-against-the-wall-anyway clarification of terminology, namely because I hate when people get things blatantly wrong.
Exit polls got the name because they survey people, in person, as they exit the voting booth on Election Day. NBC News coined the term in the 1970s, according to a 1991 account by the late Warren Mitofsky, who conducted the first modern exit poll for CBS in 1967. Before then both CBS and NBC called them "election day polls" and at one point NBC tried "street polls," Mitofsky wrote.(1)
One great strength of a true exit poll is you know you're interviewing people who actually have voted. By telephone, and perhaps to a lesser extent online (where the respondent doesn't have to talk to a stranger), the question "have you voted" may produce the civics-class answer "I'm a good citizen, of course" when that might not be the case (though with turnout as high as reported in Scotland, this problem is diminished).
Mounting in-person exit polls nationwide - with statewide polls in many states too - is quite an expensive endeavor, though. There's sound reason why the networks and AP pool resources for a single set of exit polls. And it's why some have tried alternative methods, reaching people on or just before Election Day by phone or online. (I've been involved with such experimentation myself.)
In terms of methodology, there's nothing inherently invalid about this non-exit approach, if done right. There's the problem of being sure you're interviewing an actual voter, but traditional exit polls have their challenges too. But - again, not passing judgment on the surveys themselves - "election day polls" really is a more precise name for this kind of technique.
Now, it's also true that the traditional exit poll doesn't entirely merit the name anymore. These days around one in three American voters cast ballots before Election Day, in states that to varying degrees make it easy to vote early or absentee. These voters do not go to voting precincts on Election Day so to cover them, exit polls are supplemented by telephone polls conducted the week before the election nationally and in states with higher rates of early voting. That component isn't an "exit" poll per se, of course, but "mostly exit polls" doesn't have such a great ring to it.
(1) Mitofsky, Warren. 1991. "A Short History of Exit Polls." In Polling and Presidential Election Coverage, Paul J. Lavrakas and Jack K. Holley, eds. Newbury Park, Calif.: Sage Publications Inc.
Note: Since 2012 I have managed exit polling for NBC News. I've worked with exit polls and the national media pool that conducts them since 1994, for many election cycles at The Associated Press.